Simulating molecules
Lesson 13 of 23 in the course “Quantum”. Free, with animations and a quiz.
Why the quantum computer was invented at all
The idea wasn’t born for breaking ciphers. In 1981 physicist Richard Feynman said at a conference: nature is quantum, and if you want to simulate it, do it quantumly. Electrons in a molecule live in superposition, and each extra part of the molecule doubles the number of options. By IBM’s estimate, describing a 24-atom caffeine molecule exactly takes about 10⁴⁸ bits. That’s more than the number of atoms on Earth. Qubits work the same way the molecule does, so they describe it naturally.
Drugs: key and lock
A drug works when its molecule sits in a protein’s “pocket”, like a key in a lock. Quantum effects make it hard to predict how tightly it binds, so pharma companies have to synthesise and test thousands of candidates. In May 2026 IBM, the Cleveland Clinic and Japan’s RIKEN institute modelled a 12,635-atom protein: a 156-qubit chip did part of the calculation, and the Fugaku supercomputer did the rest. It’s a size record, but the authors don’t claim superiority over classical methods and call it a starting point.
Fertiliser and millions of tonnes of CO₂
Half the people on the planet eat food grown with nitrogen fertiliser. It’s made by the Haber–Bosch process: about 450 degrees and pressure of hundreds of atmospheres. By common estimates, it consumes 1–2% of all the world’s energy. Yet bacteria do the same thing at room temperature using the enzyme nitrogenase. Exactly how isn’t fully understood: the enzyme’s core, the FeMoco cluster, is too complex for ordinary calculation. For years it was called the perfect task for a quantum computer.
The first verifiable steps
In October 2025 Google demonstrated the Quantum Echoes algorithm on its 105-qubit Willow chip. The system is run forward, one qubit gets a slight “nudge”, it’s run backward, and then you see how the “echo” changed. The calculation came out 13,000 times faster than on the Frontier supercomputer, and the result can be reproduced on another quantum computer. That matters: earlier records were nearly impossible to check. The team also tried the method on real molecules as a step toward measuring their shape.
What it looks like in practice today
Today’s chips are noisy, so chemistry is computed as a pair. The quantum processor generates candidates for how the electrons might be arranged, the part classical computers find hard. The supercomputer throws out obvious noise, refines the answer and decides what to ask the chip next. That’s how IBM’s protein work and most chemistry demos of recent years are built. IBM itself calls it quantum-centric supercomputing.
A race where both sides run
An important lesson in honesty. In March 2026 researchers showed that the very FeMoco model, the “perfect task for a quantum computer”, could be computed to chemical accuracy on classical computers by cleverly filtering the options. Classical algorithms keep improving too, and every quantum advantage has to be proven again. Real wins in chemistry are expected with error-corrected computers, which the big players promise by the end of the decade. Want into this field? Learn both chemistry and code: people at the intersection are needed.
Check yourself
- Why are molecules hard to compute on classical computers?
Correct answer: Electrons are in superposition: the options grow exponentially. Each extra part of the molecule doubles the options, just like qubits. An exact calculation for caffeine would need about 10⁴⁸ bits. - How do you honestly describe IBM’s 12,635-atom protein work?
Correct answer: A size record together with a supercomputer, with no superiority claim. It’s a hybrid calculation and a preprint. The authors themselves call it a starting point and don’t claim superiority. - What did the FeMoco example show in 2026?
Correct answer: Classical algorithms improve too, and advantage must be proven. The FeMoco model was computed on classical machines. That’s not a verdict on quantum chemistry but a reminder: the race runs on both sides.